Media access control (MAC) methods attempt to avoid collisions in wireless communication networks. Typically, the MAC specifies, schedules and manages concurrent transmissions of multiple wireless devices or nodes.
A number of MAC types are known, including carrier sense multiple access (CSMA), time division multiple access (TDMA), code division multiple access (CDMA), and frequency hopping (FH). Hybrid MAC types use a combination of CSMA, TDMA and FH. Lower layers of such a network are specified by the IEEE 802.15.4 standard, incorporated herein by reference.
The standard defines the protocol and compatible interconnection for data communication devices using low-data-rate, low-power, and low-complexity short-range radio frequency (RF) transmissions in a wireless personal area network (WPAN), such as Bluetooth. The network can include coordinator and leaf nodes in a cluster-tree like network topology.
FIG. 1 shows a superframe 160 according to the EEE 802.15.4 standard. The horizontal axis 105 indicates time. Each coordinator in the network periodically transmits a beacon 100. The beacon is used for synchronization and resource allocation. An interval between two consecutive beacons is a beacon interval 120.
The superframe includes a contention access period (CAP) 150 that uses CSMA, followed by a contention free period (CFP) that uses TDMA. The CFP 140 includes guaranteed time slots (GTS) 145. Each time slots 145 is allocated to a device that requires contention free access to the channel to minimize probability of collision of its transmission with other transmissions. Typically, the CFP is used for more important traffic that must get though in time.
The CAP 150 and the CFP 140 form the active portion 110 of the superframe 160, which is followed by a much longer inactive period 130. The inactive period can be used by other coordinators, while the coordinator device of this superframe is idle and ‘listens’ to the channel for transmissions by the other coordinators. A child coordinator 11 can start its superframe 170 during the inactive portion 130 of the superframe 160 of its parent coordinator 12. A leaf node communicates with its parent coordinator only during the active portion 110 of the superframe 160 of its parent coordinator 10. The inactive period can be several seconds.
There are several problems with the above design.
First, the active period begins with the CAP. During the CAP, transmissions are more likely to collide than during the CFP. If a transmission of a device collides with the transmission of another device, then a retransmission is necessary. In this case, the device continues competes with other devices for channel access to retransmit during the CAP 150. If the retransmissions continue to fails during the CAP, then the device has to wait until the next superframe to complete the transmission. This increases latency in the network, and tends to push traffic later out in time.
Second, it is also possible that transmission during the CFP can fail, for example, due to fast fading, attenuation, multi-path distortion and frequency mismatch. In these cases, retransmissions will have to wait for the CAP in the next superframe, which increases latency for the more important on-demand traffic. A GTS in the current superframe could be reassigned to a failed message, however, this decreases throughput for priority traffic.
Third, acknowledgements (or not) of successful transmissions are handled on a per GTS basis. That is, after each GTS is received, the receiver must switch to transmit mode to send the ACK, and the transmitter much switch to receive mode to receive the ACK, and then switch back to transmit mode again for the next GTS. Switching modes takes time, consumes scarce power resources, increases latency and reduces throughput.
Fourth, the information in the beacon is time sensitive. This means that as time passes, the resource allocation made at the beginning of the superframe is less likely to be good as time passes. This is particularly true for frequency assignments. Thus, the reliability of the information for the important and later CFP traffic is less than the reliability of the incidental earlier traffic during the earlier CAP.
Fifth, the periodicity of the beacons is relatively low, e.g., a beacon every couple of seconds. This impacts the ability of devices to synchronize and allocate resources. If a node misses a beacon, then it has to wait a relatively long time for the next beacon.
Sixth, the frequency allocations during the active interval are fixed. It is known that wireless communications are frequency sensitive. Fixed frequency allocations are less desirable.
Seventh, the current standard does not specify when child coordinators can start their superframe during the inactive period. The inactive interval essentially uses contention based access. This increases the likelihood of collisions between superframes of child nodes, and reduces the overall performance of the network.
FIG. 2 shows the frame structure for another hybrid channel access method with a beacon interval 220 between beacons 200. A superframe 260 includes a CFP-A 270, followed by a CAP 250 and then followed by a CFP-B 240 and an inactive 230 period, along a timeline 205. The contention access period is frequency hopped. The CFP-A and CFP-B include guaranteed time slots 275 as in the structure of FIG. 1. The TDMA slots are allocated to traffic with repetitive bandwidth requirements, and periodic coordinator beacons. The CAP 250 can be used for retransmissions, unscheduled alerts, and requests for TDMA bandwidths. The CFP-B 240 includes ad-hoc TDMA slots 245. These time slots 245 can be used for unscheduled bandwidth, and high priority on-demand and low latency burst mode traffic. The superframes 260 are schedules centrally. The TDMA slot allocation within a superframe 260 is managed by the coordinator, which controls the superframe. This structure only partially solves the problems associated with retransmission for the CFP-A. However, all of the other problems persist.
FIG. 3 shows a frame structure for a time synchronized mesh protocol (TSMP) along a time line 305. The TSMP is a packet-based protocol where each transmission contains a single packet and acknowledgements are generated immediately when a packet has been received. All node-to-node communications using the TSMP are transacted in a specific time slot 320.
A sequence of the time slots 320 comprises a frame 300. The frame length 330 is counted in slots, and is a configurable parameter. A TSMP node can participate in multiple frames at one time for different tasks. The TSMP does not use beacons. For each time slot, frequency hopping is used to reduce interference. The hopping sequence and slots assignments are centrally managed.
This structure solves the long latency retransmission problems associated with the structures of FIGS. 1 and 2. However, this structure requires centralized network management. This means that a local failure can shut down the entire network. Also, recovery from a failure takes a long time.